Walk Forward Schemes

Question 1

Walk-Forward Scheme

Answer 1

A rolling implementation of
* In-sample optimization
* Out-of-sample testing

Protects against the possibility of one round cross-validation that by chance we chose some input parameters that did well in both in and out-of-sample sets.

Question 2

Steps for WFS

Answer 2

• initialise training and test period
• optimisation
• test strategies
• slide window
• repeat
• aggregate results

Question 3

Initialise Training and Test Period

Answer 3

Choose an initial training period and a subsequent test period.

Question 4

Optimisation

Answer 4

Use the data in the training period to get the optimal parameters.

Question 5

Test Strategies

Answer 5

Test the strategy with the optimal parameters on the test set.

Question 6

Slide Window

Answer 6

Move the training and test periods forward in time.

Question 7

Repeat

Answer 7

Go back to optimisation and repeat the process until you’ve moved through all the available data.

Question 8

Aggregate Results

Answer 8

Collect performance metrics from each test period
to evaluate the overall performance of the strategy.
All similar - average the results/
Widely different - go and cry.